"Automated Racism": Chinese Tech Companies Profiling Ethnic Minority

By IPVM Team, Published Apr 26, 2019, 08:54am EDT

Scrutiny of Chinese facial recognition providers, including Hikvision, has sharply increased following a New York Times report that they developed analytics for the government to track members of China's Uyghur ethnic minority, "potentially ushering in a new era of automated racism."

automated racism china

Uyghurs are a mostly Muslim people in China's western Xinjiang region who are subject to a harsh Chinese government campaign of mass detentions and intrusive surveillance, a crackdown that's been lucrative for firms like Hikvision and Dahua.

The NYT article indicates a growing ethical/geopolitical divide between China and the rest of the world on video surveillance and facial recognition, threatening cooperation with China on the topics. It will likely lead to human rights questions surrounding Western investors and tech firms who continue to work with the Chinese providers named by the NYT.

In this post, we examine the report and its broader impact, including:

  • NYT Summary
  • Hikvision Mentioned
  • Comparison to West
  • Accuracy Concerns
  • Face Rec Industry Impact
  • Impact Within China

NYT ****** ******* 

*** ******* ************ ******* ********* **** developed ******** ********* "********** into *****’* ******* ********* networks ** ************ *******" which "***** *********** *** Uighurs ***** ** ***** appearance *** ***** ******* of ***** ******* *** goings *** ****** *** review."

*** ******* ***** *** following ********, ****** * anonymous ******* *** ******/********* documents:

"****** *** *****" ****** departments ****** ***** ****** to *** **** *********, with *** ********** **** a ***** ****** ****** "should ******* ****** *********** to ******** ******/***-****** **********,” and ******* ****** ************* notifying ****** ** ******** Uyghurs **** "*** **** flight ** *** **** day"

* **** ****** ******** with **** **** "**********" determined *,*** *** ** 500,000 **** **** ****** over *** ****** ** a *****.

rec uygur

********* ********* ********* **** *** tech ***** "*********** **** alarms" ** ****** "** originally *** ****** ***** in * ************, *** within ** **** *** Uighurs ******"

****** ********* **** ****** against "********** ****** ********* publicly," **** ****** ** has ****** **** ******.

Hikvision *********** 

****** *** ********* ******, NYT ******** **** ********* "offered * ******** *********** ********, but ***** ******* ** out ** ****." **** year, **** ***** *** story ***** *********'* ******** *********,***** **** ************* ********* at * ********** ** China.

***** ***** ************ ****** up ****'* ******,********* ******* **** **, quietly **-******* *** ***** and ******** ****** *******. 

Companies ********: **** ******* ** ********

***** **** *********, *** companies ****** *** ** the *** ******* **** of *****'*******-******** '********':

Comparison ** ****** ********* ** *** ****

**** ********* ****** ** the ** ** **** but *** ****. **** of *** ***** ***** surveillance ********* ***** **. However, **** **, * small (***** ** ********* according ** ********) *** Francisco-based ****** *********** ******* with *****$*** ** ******* *******, ******** *** *********** ** ****** "********* detection" ************:

ever ai

*******, ****** *** ** the ********** ** *** West ** **** **** limited ** ** ** constrained ** ********/**** ***********/***** scrutiny. ***** *** ** documented ******* ** ** police *********** ***** ****** analytics ** ****** *** suspects ** *********. *** closest **** ** **** *** Intercept************ *** **** **** video ************ ******* ** develop **** *** ******** algorithms **** **** ****** search ******** ** **** - *******, *** **** denied **** ***** *** skin **** ******* ** order ** ***** ****** profiling ***********.

Accuracy ********

*** *** ********* (*********) that **** ******* *** "imperfect" *** ****** ** a ***** ****** ** good ******** ********** *** other ************* *******, ******* the ***** ** ***** positives ******* ****** ** legal *******.

Impact ** **** *** ******** 

*** ******* ***** *** the ******* ******** ***** rights ****** ** ********** criticism ** *****'* **** recognition *******. **** *** a ****** ** *******, chiefly:

Effect ****** *****

**** ******* **** ******* led ** **** ******* in ***** - *********, for *******,*** **** ****** ******** *********** **** Xinjiang ************ **** **** Technology. *******, ***** **** SenseTime, ***** ** ** sign *** ** ******* firms ******* **** **** the ********* ******** ******** market; **** ** **** have **** ********* ** questions ***** ***** **** abetting ***** ****** ****** there.

**********

***** *** *** ******** implications, ********* ******* ********* considering ***** ******* ****** recognition ********* ********** ** active ** ******** ** developing *** **** ** minority ********* **** **** to ***** ********* **** now **.

Vote / ****

Comments (10)

Undisclosed Manufacturer #1

04/26/19 01:37pm

********* ***** *** ******** case **** *** ****** identification ** *********.  * voted *****.  **’* *** they *** **** **** should ** ******* ****.

*** ****, * ******* people ****** *** ***** to *** **** *** how **** *** **** determines **** ** ***** or ***. 

** * ******* ***** any ** ********* ** discriminate ** * *** you ***** ********* * human? **. 

***** ** ****** *** scales, **** ***** ** I ** ***!

Undisclosed Integrator #2

04/26/19 03:31pm

* ***** **** *** of *** ***** ***** manufacturers ******** ***** ****.  He **** ***** ********** is **** ****** ** protect **** **** ****.  So **** ** *****'* people *** ***** ****** by *** **********.  ** course, **** **** ** choice ****** ******* ** speak ** ******* *** government ** ** **** immediate ************ ** **** death.

*****, ****** ** ******* in *****.  ****** **** never ** **** ** it ** ***** ******* in *** *** ********** the *****.  ** **** as ***** ***** *****, fear *** ********* ****.

* ******'* ** ********* some ******** ** *** South ****** ***** ********* to **** *** ********* criminals ***** ** **** color ****. 

Undisclosed #3

In reply to Undisclosed Integrator #2 | 04/26/19 10:37pm

"* ******'* ** ********* some ******** ** *** South ****** ***** ********* to **** *** ********* criminals ***** ** **** color ****."

**** ** ******** ********** - *** *** ********* yourself ** ******* **** things.  *** ******* *** your ********* ** *** U.S. ***** **** *** Mark ***** *******.

* ** *** **** the ***** *** * have ***** **** *** decades.

**** ***** *** ***-****** who **** *** *** hats *** **** **** until **** ******* *** and ** ****... *** the ******* ********** (*****) is *** **** ****** than ***** * **** who **** ** ******** of ****** ******** ** the *****.

****:  * *** *** clear ***** ********* ** primary ******** - ***** your ******** ******** *** humanist ******** ****** **.

** ** **** *****:  I **** **** ******* to ** ******** ****** itself - ** *** apparently *** *** *.*. south ******* * ******-****** lens.

John Honovich

IPVM | In reply to Undisclosed #3 | 04/26/19 10:39pm

* ***** **'* **** to *** **** *** ethnic ** ****** ********* using ***** ************ ** the ** ***** ***** a ******** ***********.

* ** *** ****** it **** *** ****** but **** ** ****, the (***********) ******* ** it **** ** ******.

******** ** *****, ***** the ** ***** ** banned *** **** ******* and ****** ********** ******** on **** ******* *** blocked ** *****.

Undisclosed Integrator #2

In reply to Undisclosed #3 | 04/26/19 11:10pm

* ****'* *** *** in *** ***** *** racist.  *******, ** ** still ***** ********* *****, and ********** *** ***.  So ***, * *** not ********* ****** ********** by ****** *** ********* it ***** ******.

* ******, *** *** a ********?

Undisclosed #3

In reply to Undisclosed Integrator #2 | 04/26/19 11:18pm

"* ****'* *** *** in *** ***** *** racist."

******.  **** ** ****** a ******* ** **** original *********.  **** ******** was ***** *****.

"* *** *** ********* myself ********** ** ****** the ********* ** ***** happen."

*********?  ******* ***** *** out ** ** **** - ** ***** ******.  but * ** *** predicate ** ****** ********** with **** ********* ******* it ** *** ******* defense ** *** ******** that * ***** ***** that * *** *******.

Undisclosed Distributor #4

04/29/19 02:12pm

* ***** **** ********** doing ******** **** ***** will ** **** *** Americans **** *** ******** with ******...

*** ***** *** ********* and ************ **** ******** more *** **** *****.

* ******** ***** ***, early *****, * ****** MFG *** ******* ** their *** ****-******** (**** before *****'*) *** **** asked "**** ***** **** there *** *** ******?" the ****** *** "** follows *** ****** ******."

* ***'* ******* **** kind ** *****.

Undisclosed Integrator #5

04/29/19 09:27pm

** **** ******** - Should ******/****** ********* ** legal? * *** ***. Why - ** ***** is * **** *** for * **** ******* red ****; *** **** that * ***** ** the ****** ** **** skinned ** ******.

** ** *** *** fact **** ***** ** an ********* **** *** analyze ***** ****, ** is *** *** ******** is ****. * **** this ****** **** - Guns ***'* **** ******, people **** ******." ********** aren't ******, *** ********* could ** **** ** a ****** ******. ********** need ** ** ********* to ** ** ******** as ******** ***** ** all *** ******** *************** of * ******. 

** **** **** ***'* be **** ** * characteristic, **** *** ** used? ****, ******** **** answer, ******** ** ***** to *** *** *** biased. ******, ******, ******, skin ****, **** *****, hair ****** *** *** perceptions. *** *******'* ****** the *** ** *** perception.

*** ****** ****** *** use ** ** ********* in * ****** ******. In **** *****, * preemptive *** ** **** an ********* ***** ** biased; * **** ** know **** * ******* type ** ****** ****** the ************.  ***** ******** use ** *** **** algorithm **** *** **** parameters ***** ** **********; a *' ****, ***** set, ***** ****, ****** aged ********* *** *****. 

Undisclosed Manufacturer #6

In reply to Undisclosed Integrator #5 | 04/30/19 12:27pm

********** ****'* ******, *** algorithm ***** ** **** in * ****** ******.

**** *** ********** ******** with ********** ***** ******, so **** **** "*********** bias" ** * **** term *** ******* *** data ********** ** *****.

*****://***********.***/****-******/****/**/***-**********-***-**-******-*****-***/

Undisclosed Manufacturer #1

04/30/19 05:12pm

********** ***** ** ******* if ****** ****’* ****** them. 

Read this IPVM report for free.

This article is part of IPVM's 6,736 reports, 909 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

Already a member? Login here | Join now
Loading Related Reports